I’m having a difficult time reviewing this paper because I’m not entirely sure what I should say.  It is the first paper to compare two models using multi-latitude control in SAI, which is a great idea.  But a lot of the analysis, results, and conclusions are basically a form of “different things are different”.  I don’t know what the research community is supposed to do with that.  The authors say at the end that more models should do this experiment, but that doesn’t seem like useful effort if the best we’ll be able to do is that we’d just get even more different results.

There is a goldmine of information in the simulations the authors analyze.  I would encourage them to dig into mechanisms and provide some insights.  If we know why the models are giving us different answers, that can tell us about what the most important uncertainties might be.  As an example, the authors say on line 136 that the three factors are expected to be different in UKESM1.  So why not look at them?  On line 210 you talk about the hydrologic sensitivities being similar or different depending on the context/simulation.  So what are they, and why are they different?  When you say “likely” on line 241, can you do better?  I’m not listing every example in the paper, but hopefully the authors see what I mean.

I also think the authors are missing an opportunity with the multi-model comparison to talk about what this study could mean.  If multiple plausible models are getting a different range of achievable climates, that’s really useful to know in terms of the limitations of SAI (or at least SAI with a four latitude strategy).

I noticed there is a lot of emphasis on Arctic amplification, especially differences in Arctic amplification between the two models.  But SAI does a pretty good job of reducing Arctic warming in both models.  So I have to wonder how important model differences in AA actually are.  If you did some quantification of feedbacks or mechanisms, we might be able to figure out how important AA is.

A few more comments:

Figure 1c:  It looks like you have two different timescales of response here, in that there’s a clear shift around 2045.  Kravitz et al. (2016) found something similar, which may suggest some mechanisms that you could pursue to explain this behavior.  Relatedly, I acknowledge that the controller is prioritizing T1 over T2, so I would expect deviations in T2 from the reference value.  But why is T1 not controlled very well?

Line 26:  some missing words

Section 2:  Be more specific about what the temperature targets are and the control gains.

This paper makes the first comparison of two models implementing the same feedback control style implementation of SAI, and as such it makes a major contribution to the literature. The study compares the realized temperatures against the target temperatures for the models, finding that UKESM1 misses these targets due to substantial arctic amplification, and compares the injection strategies needed and resultant aerosol burdens, noting substantial differences, including a large southern-hemisphere bias in CESM2. It also compares the hydrological response of the two models, finding substantial differences between the models and significant shifts in tropical precipitation in both models. There is some useful discussion of the potential reasons behind the differences in the aerosol distribution and failure of UKESM1 to achieve the temperature target, though there could have been more discussion of the implications of the inter-model differences in the experiment-as-realized for the design of future feedback control experiments intended for intermodel comparisons.

I’d recommend the article for publication after the following, fairly minor comments are addressed.

The introduction gives a fine introduction to the topic, the history of the field and of modelling efforts. However, it doesn’t devote much text at all to reviewing the climate response seen in previous studies or to setting up theoretical expectations about what we may expect from the comparison. Would be nice to see a bit more of this up front.

The results (and discussion) section generally does a good job of explaining the findings and putting them in the context of previous results, but there could be more of this for the hydrological results. This could be quite simple in that climate models are notorious for disagreeing on regional hydrological change, and so noting that the fact that these models disagree here is hardly surprising.

The conclusion section could benefit from some discussion of the fact that while the models followed the same experimental procedure and so nominally are simulating the same experiment, they are in effect quite different experiments-as-realized. That is, they produce different patterns of temperature change, driven by different patterns of forcing (and they have a different baseline, though this seems less significant). To what extent are the authors comparing like with like? What are the implications of this? What changes in setup could produce experimental designs that would be more suitable for a future model intercomparison or are such profound differences in experiment-as-realized inevitable? If so, what does this imply for the assessment of SAI?

Specific comments

L30 – I’d suggest sticking to one of SRM or solar geoengineering or solar climate intervention and using throughout.

L40 – non-synergistic – another term?

L45 – “consistently” instead of commonly?

L56-60 – “ensemble” seems a bit grand to describe a pair of models.

L66 – spare parentheses?

L63-79 – Would be good to mention the ocean models too.

L93 – what longitude? into a single gridcell?

L101 – if these definitions are short, just include them here or describe the equation.

L103-110 – It would be good to mention what CESM2’s temperature increase relative to its preindustrial is. And to have some discussion of the consequences of this appear at an appropriate point in the paper.

L112 - “Results and discussion” would be more accurate.

L125-155 – I found this difficult to follow. It would be clearer to present the figure first and then compare that to other studies, or to not mention the figure until you wish to describe it. Also in describing the results of Richter et al. 2022 are you also describing Figure 2b?

Figure 2 – Hot / Cold colours or similar to indicate north vs. south would make it clearer that Arise is doing something surprising in putting so much in one hemisphere. Some text to this effect would also be good, e.g., Reporting the percentage in NH vs SH would make this more obvious to the reader rather than just stating that it “required more” in the SH.

L165 – perhaps make the change of focus clearer, by saying: “this reduced lifetime is compensated by the injection being 50% greater” or something to that effect.

L163-171 – Here as well a quantitative comparison of NH vs SH burden or AOD would be nice, e.g. in “CESM2 75% of the SO4 is in the SH”

L169 – is this comparison backwards? Doesn’t the SO4 cause the AOD increase?

L170 - “not shown” or provide some citation.

Figure 4, 6, and 7 – These map plots all have overly long titles (cut dates?) and too much white space. The hydrological maps are hard to read. Try to make these map plots as large as possible.

L196 – and a smaller decrease in Arise. A recovery suggests time has passed.

L197 – Wouldn’t this blob have a greater effect on T1 and T2 than global?

L209-216 – I’d suggest considering reporting results in %/C as this is how I understand hydrological sensitivity and as the models show quite different temperature responses. I’d also need more convincing that the pattern of temperature is the main driver, rather than the substantially different magnitude of cooling. Perhaps some comparison to the literature would help make the case.

Figure 5 – I’m presuming 1.5C here means “the baseline” Which is defined differently in the two models and doesn’t correspond to the model’s version of 1.5C in the case of CESM2. I’d suggest switching to just temperature anomaly.

L218-225 – As previous sections cited existing literature it’s probably worth noting that regional differences in precipitation between models is common.

L226-227 – Not sure exactly what this last line means, e.g., the uncertainty in the precip response due to GHGs includes uncertainty in its effect on temperature, whereas here that temperature-based uncertainty is eliminated. The fast, forcing-driven effects are combined in Arise, but the slow-temperature driven ones are eliminated.

L241-243 – How is the causality being determined here?

Figure 8 – Would seem more useful for this to appear directly before or after the figure on mean precipitation.

L256 – perhaps clarify what exacerbation means in this context.

L265-end – Given the length of these bullet points, I’d suggest reframing as normal prose.

L274-275 – This isn’t particularly clear.

L281-289 – Seems odd to have the injection last.